Environment and Planning A 2002, volume 34, pages 191- 216
DOI:10.1068/a34119
Location of adult children as an attraction for black and
white elderly primary migrants in the United States
Kao-Lee Liaw School of Geography and Geology, McMaster University, Hamilton, Ontario L8S 4Kl, Canada; e-mail: [email protected]
.
William H Fre}1T Population Studies Center, University of Michigan, Ann Arbor, MI 48104, USA; e-mail: [email protected] Ji-Ping Lin The Institute of Labor Studies, National Chung-cheng University, Ming-hsiung, Chiayi, Taiwan 621; e-mail: [email protected] Received 11 June 2001; in revised form 14 June 2001 Abstract. By using infonnation on state of birth in the 1990 Census of the United States, the authors
create a variable serving as a proxy for the distribution of adult children in 1985. This variable is then
used in a two-level nested logit model to explain the 1985 - 90 interstate primary migration of elderly
black people and elderly white people within the context of environmental amenities and other
factors. The main fmdings are as follows. First, the location of adult children as well as environmental
amenities are among the most important· attractions for the primary migration of elderly black and
white people. Their effects are stronger on white people than on black people. Because a relatively
high proportion of out-migrated black adult children are located in the industrial states of the
snowbelt and a relatively high proportion of their white counterparts are located in high-amenity
states of the sunbelt, the attractions of adult children and environmental amenities are more prone to
counter each other for elderly black people and to reinforce each other for elderly white people.
Consequently, the net transfers of elderly primary migrants are small and somewhat oriented toward
the snowbelt for black people, but they are voluminous and strongly oriented toward the sun belt for
white people. Second, the attraction of adult children is strong for not only the unmarried but also the
married elderly, although it is somewhat stronger for the widowed aged 75 years and over. This
finding can be taken as evidence for the viability of the 'modified extended family' system, which
not only legitimizes the out-migration of adult children for career advancement but also encourages
the migration of elderly parents to be close to their non-coresident children for services that require
continual proximity. It also suggests that the elderly do not have a strong tendency to delay their
migration toward non-coresident children until the loss of a spouse or becoming very old.
1 Introduction
The idea that the location of adult children can be influential in attracting elderly
migrants in an industrialized country such as the United States can be traced back to
the seminal paper of Litwak (l960b), which provided part of the basis for the develop­
ment of the theory of the 'modified extended family' (MEF) (Litwak, 1965; 1985).
According to this theory, the MEF is better than the classical extended family and
the isolated nuclear family because it not only legitimizes the out-migration of adult
children for career advancement but also encourages the relocation of elderly parents to
be near their adult children for the services that require continual proximity. Among the
various important implications of this theory is the one on the provision of assistance
and services to the elderly (Litwak, 1985). Thus, an empirical study on the attraction of
elderly migrants to adult children is important both for theoretical and for practical
reasons. Unfortunately, it has been difficult to carry out such a study because of the lack
,-] Also at the Milken Institute, Santa Monica, CA.
,
.
K-L Liaw, W H Frey, J-P Lin
192
of empirical data that contain information on the locations of non-coresident children
(Clark and Wolf, 1992).
The pattern of the net transfers of elderly black 'primary migrants' (that is, migrants
leaving their state of birth) from sunbelt states to snowbelt states with a large working­
age black population (Frey et ai, 2000) suggests that the location of adult children is
essential for explaining the migration behavior of elderly black people. The fact that
widowed elderly white people, relative to married elderly white people, are more likely
to undertake primary migration from northern industrial states (for example, New York
and Illinois) to California (where a high proportion of their out-migrated adult children
are concentrated) also suggests that white adult children are more likely to welcome
their elderly parents to move closer to them when their parents' need for assistance
becomes greater. (I)
Our main purpose in this paper is to assess the importance of the location of adult
children in influencing the 1985 - 90 interstate primary migration of elderly black
people and white people in a multivariate context, based on the data of the 1990
Census. As the census questionnaire did not elicit information about the location of
non-coresident children, our empirical work will first show that it is possible to use
information on the state of birth to create a reasonable proxy for the location of the
adult children of elderly 'natives' (that is, the elderly whose 1985 state of residence was
identical to their state of birth). This proxy, together with indicators of environmental
amenities and other explanatory variables, will then be used in a nested logit model to
account for the primary migration of these people. It is the comparison of the effects
of the adult children's location, on the one hand, and environmental amenities, on the
other, that will help us account for the major differences between black and white
elderly migration patterns.
In addition to yielding more insight into the differences between black and white
elderly migration, our findings will shed further light on the viability of Litwak's model
of the MEF, which he claimed to be the only one that is consistent with modern
industrialized society (1985, page 102). Moreover, they will also help us explain an
'unexpected' contrast between return and nonreturn in-migrants of the midwest that
was revealed by Longino and Serow (1992) but has to date remained unexplained:
relative to return in-migrants, nonreturn in-migrants are older, more likely to be
widowed, and less likely to live independently.
Our focus on primary elderly migrants is based on the well-known classification of
primary migrants, onward migrants (that is, those who migrate from a nonbirth state
to another nonbirth state) and return migrants (that is, those who migrate back to their
state of birth) that has been used in previous census-based migration studies (Eldridge,
1965; Long, 1988; Longino, 1979; Miller, 1977; Newbold, 1996; Rogers, 1990; Serow,
1978). This classification is essential for the creation of reasonable proxies for the
location of adult children from census data. Owing to space limitations, the results
of our assessment on the attraction of onward and return elderly migrants to the
location of their adult children, which reinforce the main conclusions of our analysis
on elderly primary migration, will be reported in a separate paper. For elderly white
people, the 1985 90 net transfers of migrants between major gaining and losing states
Our examination of the 1990 Census data reveals that 1985-90 primary out-migration rates
from New York to California are 0.26% for widowed elderly white people and 0.23% for married
elderly white people. The corresponding figures for those from Dlinois to California are 0.33% and
0.29%, respectively. By 1985, 5.8% (7.0%) of the adult children born in New York (Illinois) had
relocated to California-their favorite destination. Their birth·to·1985 stayer rates are 60.4% for
New York and 58.8% for lIlinois. If the attraction of adult children is ignored it is difficult to
explain why the amenity-rich state of California should be more attractive to the widowed elderly
than to the married elderly.
(I)
;:
\
.
,.
Adult children and elderly migration
193
were greater for primary migration than for onward and return migration. For elderly
black people, these net transfers were greater for return migration than for onward and
primary migration (Frey et al, 2000, table 4).
There are in our study methodological and theoretical reasons for making the
distinction between black people and white people. The methodological reason is
that the patterns of lifetime (that is, birth to 1985) migration of adult children differ
markedly between the two. In major northern industrial states, white adult children
were much more likely to out-migrate and to select amenity-rich destinations than
were black adult children. By contrast, in the southern states with a large black
popUlation, black adult children were much more likely to out-migrate and to head
for northern industrial states than were white adult childrenP) To the extent that the
attraction of the out-migrated adult children is influential, black and white elderly
migrants are subject to pulls in opposite directions. Consequently, failure to keep the
black - white distinction will cause the opposite pulls to cancel each other out and
result in a serious understatement of the importance of the attraction of adult children.
The theoretical reason for keeping the black - white distinction is discussed in the next
section.
2 Literature review and research context
Our search through the literature revealed that none of the previous multivariate
analyses of elderly migration (for example, Clark et aI, 1996; Freyet aI, 2000; Newbold,
1996) used the location or distribution of adult children to explain elderly migration.
The reason for this seems to be that there has not been any national data set that
identifies the specific location of non-coresident adult children. Two of the nationally
representative longitudinal data sets that have been used in recent years for obtaining
valuable insight into US families and for studying the relocation of the elderly in the
United States [the Longitudinal Study of Agin.g (LSOA), and the National Survey of
Families and Households (NSFH)] contain information on the proximity between
elderly parents and their children (measured in terms of travel time in the LSOA and
in terms of travel distance in the NSFH)as the only geographical information for non­
coresident adult children. Based on these data, some knowledge about the attraction of
elderly migrants to their adult children may be only indirectly inferred from changes in
proximity or from transitions into and out of coresidence. Because changes in the
location of non-coresident adult children were not recorded it is not possible directly
to attribute changes in parent-child proximity to the migration of one or the other.
From the LSOA, Silverstein (1995) found that for noninstitutional persons who
were aged 70 years and over in 1984 and had at least one surviving child in 1985
(N = 3468), the propensity to become temporally closer to their adult children
between 1984 and 1988 was enhanced by a recent decline in their physical health. He
also found that "the conjunction of declining health and widowhood increased both
the degree of non-coresident proximity and the likelihood of transition to coresidence"
(page 29, emphasis in the original). To detect the effects of 'race' and ethnicity, he used
a single dummy variable to represent black and Hispanic people in his multivariate
models and found that this variable did not have a statistically significant effect either
on the propensity to converge (to become closer to a child) or on the degree of
convergence, but it did have a significant negative effect on the propensity to diverge.
Among the white adult children born in New York, 40% had out-migrated, among whom 14%
ended up in Florida by 1985. The corresponding figures for the black adult children born in New
York are 26% and 6%, respectively. Among the black adult children born in Mississippi, as many
as 64% had out-migrated among whom 29% ended up in Illinois by 1985. The corresponding
figures for the white adult children born in Mississippi are 41% and 3%, respectively.
(2)
\
.
194
K-L Liaw, W H Frey, J-P Lin
The negative effect suggests that, relative to elderly white people, elderly black and
Hispanic people are less likely to make amenity-oriented interstate migrations that
tend to increase the average distance from their children.
By using the longitudinal data of the NSFH, Rogerson et al found that for the
respondents who were aged 60 years and over at the initial survey in 1987 - 88 and had
at least one surviving child with a valid distance measure at both the initial survey and
the second survey in 1992-93 (N = 1285), "an increase in functional limitations is the
most consistent predictor of geographical convergence between elderly parents and
their adult children" and that "the onset of widowhood during the observation period
leads to a greater likelihood of living with an adult child" (1997, page 121). Rogerson
et aI, by using in their multivariate models a single dummy variable to represent black
people and those in all other minorities, also found that this dummy variable does not
have any statistically significant effect on (1) the odds of convergence compared with
no change, (2) the odds of divergence compared with no change, (3) the odds of living
independently compared with no change, and (4) the odds of living jointly compared
with no change. It is likely that the complete lack of statistically significant effect of
this dummy variable is partly a result of the smallness of the sample size and the
grouping of black people with those in other minorities.
By using the cross-sectional data of the 1987 88 NSFH, Clark and Wolf (1992)
. studied the effect of migration of the elderly (aged 60 years and over, with at least one
child aged 19 years or over; N = 2714) on the proximity to their children in a multi­
variate framework. They defined migrants as those who had moved more than 25 miles
in the five years before the survey. They also represented proximity by a dummy variable
that assumed the value of I if the parent in question coresided with a child or if the
distance to a child was within 10 miles at the time of the survey. They found that
the effects of elderly migration on proximity to children had a curvilinear age pattern:
"Among the 'young old', migrants are less likely to live near a child than are non­
migrants. However, the older migrants are, the more likely it is that they live near at
least one child. By age 77, migrants are more likely than non-migrants to be in
close proximity to a child" (Clark and Wolf, 1992, page 87).0)
They also found that widowed respondents were more likely to live near a child than
were those of other marital statuses but that functionally limited older parents were no
more and no less likely to live near a child than were other parents. Clark and Wolf did
not attempt to examine the potential effect of 'race' on parent-child proximity.
Another multivariate study of the same data by Lin and Rogerson (1995) showed
that elderly black people and white people did not differ significantly in the proximity
to their closest and second-closest children.
Although the effects of 'race' and ethnicity on parent-child proximity and on its
change appear to be largely nonsignificant, LSOA and NSFH data provided clear
evidence that increase in functional disability significantly increases the elderly's
proximity to their children and that this effect is reinforced by becoming widowed.
Is this increase in proximity achieved mainly by the movement of the elderly or by
the movement of their children? Speare and McNalley'S (1992) analysis of the data
of the Survey of Income and Program Participation indicated that more than two
thirds of elderly parents who became geographically closer to their children did not
move themselves, suggesting that the increase in proximity was mainly attributable to
(3) The finding that, among the 'young old', migrants are less likely to live near a child than are
nonmigrants mayor may not mean that the greater parent-child distance of the migrants are a
result of the migration of the elderly. It may be mainly a result of the previous out-migration of
their adult children. For an interesting discussion and some empirical evidence of this possibility,
see Bultena and Marshall (1970).
\
.
Adult children and elderly migration
195
the movement of their children. However, an analysis of LSOA data by Bradsher
et al (1992) showed that the elderly's propensity to change residence was enhanced by
increase in instrumental disability and that this enhancing effect was particularly
strong for the recently widowed. Together with the above-mentioned finding of Clark
and Wolf (1992), this finding suggests that migration of 'weakened' elderly may have
contributed significantly to the reduction in distance from their children.
What remains unclear in the empirical studies is the strength of the attraction of
healthy and married elderly to their adult children. There are several reasons for
expecting that it can be quite strong. First, a non-coresident child may reside in an
amenity-rich region and provide information and help to facilitate the migration of her
or his healthy and married parents to that region. This type of migration is likely to
happen, because migration can enable the elderly to enjoy not only environmental
amenities but also visits with their children. Second, some healthy elderly couples
may move to the vicinity of their children in order to maintain exchanges of services
and affections, including interactions with grandchildren. Third, some elderly may feel
safer by moving closer to their children before they experience a decline in health or
the loss of spouse. For these reasons, our study will include the elderly of all marital
statuses and different ages. We use age as a crude proxy for health status.
It is common for researchers of elderly migration to put their studies in the
context of the three-stage developmental framework of Litwak and Longino (1987):
(I) amenity-oriented migrations, mostly by relatively young, healthy, and married
couples; (2) migrations of the partially disabled or widowed elderly toward their
adult children or other kin; and (3) movements of the seriously disabled elderly
into institutions. However, we think that it is more meaningful to put our study in
the context of the theory of the MEF (Litwak, 1960a; 1960b; 1985). The theory is well
grounded in empirical evidence and generates a relatively optimistic prospect for the
future of older populations in today's industrialized society. (4)
The MEF consists of "a series of nuclear household units that are semi-indepen­
dent of each other" (Litwak, 1985, pages 101 - 102). In contrast to the classical extended
family (CEF) that discourages differential geographical and occupational mobilities,
the MEF legitimizes them (Litwak, 1960a; 1960b). Because such differential mobilities
are essential for the advancement of family members in the formal institutions of an
industrialized society as well as for enhancing the productivity of the modern economy,
the MEF is consistent with the basic nature of modern society whereas the CEF is not.
Despite being prone to be separated by geographical location and social status, most
members of the MEF are able to maintain substantial exchanges of services and
affections among themselves by using modern technology (for example, telephones,
cars, airplanes, and money in the banking system).
An important challenge for the MEF occurs when some member (usually an
elderly parent) experiences a long-term disability that requires the continual proximity
of another non-coresident member who is willing or obliged to provide daily instru­
mental assistance. As an elderly parent who has retired from the formal institution of
employment is free from job-related mobility constraints, it is likely that she or he may
(4) By the 1940s the idea that in industrial societies the classical extended family system is bound to
be replaced by the isolated nuclear family system was generally taken for granted by social
scientists, including major social theoreticians such as Wirth (1938) and Parsons (1949). However,
since the 1950s empirical evidence showing the continuing strength of the connections between
elderly parents and their adult children started to accumulate in England (for example, Townsend,
1957; Young and Willmott, 1957) and the United States (for example, Litwak, 1960a; 1960b). For
more evidence, see Sussman (1965, page 66), who mounted one of the most forceful attacks on the
"isolated nuclear family myth", a 'myth' that can be partly traced back to such great European
social thinkers as Durkheim and Weber.
I.
•
K-L Liaw, W H Frey, J-P Lin
196
migrate to the vicinity of an adult child soon after or even before the onset of long-term
disability. The assessment of this likelihood is essential for judging the viability of the
MEF as a humane subsystem in an aged society. A negative result of this assessment
would imply a high risk that the family system may degenerate into the isolated nuclear
family system whereby the elderly with long-term disability can expect little instrumen­
tal assistance from their children. It would then imply the need for the large-scaled
proliferation of formal institutions for the elderly as a practical way of dealing with the
aging trend.
In a modern society, the main difference between members of the MEF, on the
one hand, and friends and neighbors, on the other, as helpers for the elderly is that
members of the MEF can maintain long-term commitments whereas the friends and
neighbors usually do not. Mainly because of this difference, friends and neighbors
usually cannot be a suitable substitute for members of the MEF when the elderly
need long-term instrumental assistance (Litwak, 1985). There is empirical evidence
that long-term commitments among family members and the Willingness to engage in
intergenerational transfers are better developed in households with a relatively stable
marriage and children living with both parents (Pezzen and Schone, 1999; Silverstein
and Bengston, 1997). In an innovative study of black and white families based on the
microdata of the 1880, 1910, 1940, 1960, and 1980 Censuses of the United States,
Ruggles (1994) identified a fundamental and persistent difference between black and
white households: the former, though more likely to be extended, have been more likely
to have young children without one or both coresiding parents. This difference is
related to greater proportions of nonmaritaI births, marital instability, and desertion
by fathers among black people (Morgan et aI, 1993; Smith et aI, 1996), which are in
turn related to the legacy of slavery and the ongoing marginalization of the majority of
black people in the broader socioeconomic context (Biblarz and Raftery, 1999; Frazier,
1939; Jaynes and Williams, 1989;Lichter and Eggebeen, 1993; Moynihan, 1965). Conse­
quently, it is likely that the structure of black families has been less conducive to the
development of long-term commitments, which are essential to the proper functioning
of an MEF.(5) In a study based on the 1987 - 88 NSFH data, it was found that "African­
Americans are consistently less likely than whites to be involved in interg.enerational
assistance" (Hogan et aI, 1993, page 1428). Thus, both the theory of the MEF and some
empirical evidence lead us to expect that elderly black people are less likely to be
attracted to the location of adult children than are elderly white people.
3 Data and statistical model This research uses the 8% data from the 1990 Census PUMS (public use micro­ sample) files: the 5% State PUMS files combined with the 3% PUMS-O files. By comparing the state of residence in 1985 and the state of birth, each US-born individ­
ual who resided in the United States in 1985 is identified as either a (same-state) 'native' or a 'nonnative'. The sample used in this study of primary migration includes Ruggles (1994) also found that although extended households have been more common for black
people than for white people, elderly black people (aged 65 years and over) were less likely to live
with their own children than were elderly white people until 1940. To the extent that coresidence
reflected long-term commitments by family members before the sharp rise in income and social
security allowed many elderly Americans to choose independent living arrangements after World
War 2, his finding suggests that for a long time elderly black people probably received less help
from their adult children than did elderly white people. As a consequence of the very sharp decline
in the rate of coresidence of elderly white people with their children in recent decades, elderly black
people have become more prone to coreside with their children than are elderly white people,
although their rate of coresidence had also declined (from 44.8% in 1940, to 32.9% in 1960, and
22.8% in 1980).
(5)
\
.
197
Adult children and elderly migration
all black and (non-Hispanic) white natives who were aged 60 years and over in 1990.
Among elderly natives, 'primary migrants' are defined as those whose 1985 and 1990
states of residence were different, with the remaining individuals defined as 'stayers'.
In order to retain information on key personal factors and to make the input data
files for the statistical model into manageable sizes, the sample weights of all black and
white elderly natives are used to create a multidimensional tabulation. The dimensions
of the tabulation include: (I) 'race' (black, white); (2) educational attainment (less than
high school, high-school graduation, some college ·education, college graduation); (3)
marital status (single, married, widowed, divorced, separated); (4) age in five-year
groups (60-64, ... , 80-84, >85 years); (5) gender (female, male); (6) poverty status
(poor, nonpoor, unknown); (7) state of residence in 1985; and (8) state of residence in
1990. Poverty status is defined according to the official poverty line. Only about 4% of
the elderly natives had an unknown poverty status.
Our multivariate statistical model is a nested logit model that is based on a two­
level choice framework: a choice between departing and staying put at the upper level,
and the choice of a specific destination at the lower level. For an elderly native with
personal attributes s and residing in state i in 1985, we specify that her or his
migration behavior in 1985-90 depends on a departure probability P(s, i) at the
upper level, and a set of destination choice probabilities, P(j Is, i) for all j not equal
to i, at the lower level. By assuming that the elderly native makes the migration
decision by maximizing her or his quality of life, these probabilities can be derived
as functions of observable explanatory variables in the following two submodels
(Kanaroglou et aI, 1986).
3.1 Destination choice submodel
exp[b T xU, i, s)]
Lexp[bTx(k, i, s)]'
PU Ii, s)
j f i,
(1)
kid
where xU, i, s) is a column vector of observable explanatory variables, and b T is a row
vector of unknown coefficients.
3.2 Departure submodel
P(i, s)
=
exp[d + fTy(i, s) + u/(i, s)] ,
1+ exp[d + fTy(i, s) + u/(i, s)]
(2)
where y(i, s) is another column vector of observable explanatory variables; fT is a row
vector of unknown coefficients; d and u are unknown coefficients, with u being
bounded between 0 and 1; and /(i, s) is the so-called inclusive variable defined as:
/(i, s)
In
L exp[b
T x(k,
i, s)].
(3)
bfi
The inclusive variable represents the attractiveness of the rest of the system perceived
by the potential migrant in state i.
Assuming that the migration behavior of all persons in the same cell of the multi­
dimensional tabulation depend on the same set of P(i, s) and P(j Ii, s), we estimate
the unknown coefficients in equations (1) and (2) sequentially by the maximum
quasi-likelihood method (Liaw and Ledent, 1987; McCullagh, 1983). In constructing a
relatively concise specification of a submodel (to be called the best specification for
simplicity), we only include the explanatory variables that are statistically significant
\
.
K-L Liaw, W H Frey, J-P Lin
198
(that is, those whose t-ratios have a magnitude of at least 2.0) and that are substantively
sensible. (6)
The goodness of fit of a given specification of a submodel is to be measured by the
p 2-statistic, where
P2
x
=
L
1 -.....!.
Lo '
(4)
where Lg is the maximum quasi-log-likelihood 'of the given specification, and Lo is the
maximum quasi-log-likelihood of the corresponding null submodel (that is, the desti­
nation-choice submodel with b T = 0 or the departure submodel with IT = O. Note
that the ceiling of p2 is much less than 1.0 so that a value of 0.2 may indicate a very
good fit (McFadden, 1974).
To help evaluate the relative importance of one subset of explanatory variables (say,
the variables representing the attraction of adult children) against another subset
(say, the variables representing environmental amenities) we will delete the two subsets
of variables in turn from the best specification and then compare the resulting decreases
in p2: the greater the decrease, the more important thel deleted subset of variables.(7)
The decrease in p2 arising from the deletion of a subset of explanatory variables is
termed the marginal contribution in p 2.
4 The specification of explanatory variables
The explanatory variable at the focus of this study is 'adult children', which is defined
in the following way. For a given 'race', consider a group of elderly natives (aged 60
years and over in 1990) who in 1985 resided in state i, which is by definition also their
state of birth. Where could their adult children be located at the beginning of the
1985 - 90 migration interval? It is likely that most of their children were born in state
i and were aged 30 - 59 in 1990. The distribution of these children at the beginning of
the 1985 - 90 migration interval can be estimated reasonably well by examining the
'race'-specific birth-to-1985 out-migration pattern of the 30-59 age-group from state i.
Let C(r, I) be the number of individuals in the 30-59 age-group who were born in
state i and of 'race' group r. Also, let C(r, i,}) be the number of individuals in C(r, I)
who made the birth-to-1985 migration from state i to state j, and let C(r, i, I) be the
number of individuals in C(r, I) who remained in state i in 1985. The variable 'adult
children' is then defined as
c (r,
..) =
I,}
C(r, i, j) 1000/
C(r, i)
10
(5)
The (-ratio associated with the coefficient of an explanatory variable is computed by dividing the
estimated coefficient by its asymptotic standard error. For logit models, the maximum quasi­
likelihood (MQL) method yields the same estimated coefficient as does the maximum likelihood
(ML) method. An advantage of the MQL method over the ML method is that the MQL method
does not depend on the unrealistic assumption that migrations are independent events, whereas the
ML method does. A consequence of this difference for our data is that the asymptotic standard
errors for computing the (-ratios are much larger in magnitude than those generated by the ML
method. If we were to use the ML method we would have the difficulty of being unable to remove
variables that contribute almost nothing to the model's explanatory power.
(7) Another criterion for comparing the relative importance of two subsets of explanatory variables
is the P-vaIue, computed from the corresponding changes in the X2-statistic and the associated
degrees of freedom. However, in our model, all the changes in / are so big that the / distribution
function in languages such as SAS and QUAITRO yields either a zero or an error message for the
P-values of all deleted subsets of variables, making the comparison impossible. Our experiences in
other studies where the P-vaIues are computable show that the ranking by decreases in p2 and the
ranking by P-values are very similar.
(6)
,
.
Adult children and elderly migration
199
in the destination-choice submodel, and as
c(r, i, i)
qr, I,. i) 100%
qr, l)
(6)
in the departure submodel. It is approximately correct to say that c(r, i,}) is the
proportion of adult children born in state i who migrated to state j sometime between
birth and 1985 and who remained in state j in 1985 (the beginning of the 1985 - 90 time
interval for studying elderly migration) whereas c(r, i, z) is the proportion of adult
children born in state i who remained in state i in 1985. We expect that c(r, i,}) should
have a positive coefficient in the destination-choice submodel (implying that elderly
migrants are more likely to be attracted to a potential destination where a higher
proportion of their adult children are located) whereas c(r, t, i) should have a negative
coefficient in the departure submodel (implying that elderly potential migrants are less
likely to depart from a native state where a higher proportion of their adult children
have remained). As the effects of the variable 'adult children' are expected to vary by
'race', marital status, and age we will also create interaction variables by multiplying
this variable by dummy variables representing these personal attributes.
We represent environmental amenities by coldness of winter, cloudiness, and 'Gold
Coast', defined in the following way.
Coldness o/winter: for each state, this variable is defined as a weighted average of the
heating degree-days of cities with records from 1951 to 1980, using city populations as
the weights. The unit is 1000 degree(F)-days.(8)
Cloudiness: this is the weighted average of the number of cloudy days in a year of the
cities within a state, with the weights being the popUlation sizes of the metropolitan
areas where the cities are located. The unit is in 10 days (that is, the value of 90 cloudy
days is coded as 9 for this variable).
Gold Coast: this is a dummy variable assuming the value of I if the state in question is
on the Atlantic coast between Virginia and Florida or one of the three states on the
Pacific Coast. In the context of the above two amenity variables, this variable is used
to represent the attractions of water, mountains, and scenic beauty (Longino, 1995,
page 18).
As it is expected that amenity-oriented migrations are more likely to be made by
the elderly who are relatively young (recently retired), well educated, white, and
married (Haas and Serow, 1993), some of these place attributes may have some
significant interactions with the dummy variables representing the distinctions in
personal factors such as age, education, 'race', and marital and poverty status.
Our assessment of the importance of the attraction of adult children and envi­
ronmental amenities are performed in the context of a set of other place attributes
that are considered as covariates. These covariates represent cost of living, generosity
of Medicare and Medicaid programs, proportion of homeownership, rate of violent
crime, racial similarity, relative location between origin and potential destination,
economic conditions, and the size of ecumene (that is, population size). To maintain
The data source for heating degree-days and cloudy days is the US National Oceanic and
Atmospheric Administration. The reference value for computing the coldness measure is 65°(F).
The measure is the sum of the deviations of daily mean temperatures below this reference
value. The summation is over all 365 days of a year. If the mean temperature on a given day is
500, then the contribution of this day to the sum is 150. If the mean temperature on another day
is 80 then the contribution of that day is zero. Because most days with positive values are in
winter, the coldness measure is called 'coldness of winter' for ease of communicating the main idea.
Because we use 1000 degree-days as the unit, the sum of 3500 is coded as 3.5 for this variable.
(8)
0
,
\
.
K-L Liaw, W H Frey, J-P Lin
200
the flow of the paper, their operationa1 definitions are relegated to the appendix. To
achieve a high level of explanation and to be consistent with theory, these place
attributes are also used to form interaction terms with personal factors. For example,
we use the interaction between the log of distance and the dummy variable represent­
ing postsecondary education in the destination choice submodel to allow the weaker
distance-decay effect on those with higher educational attainment.
S Empirical findings
5.1 Destination choices
The sharp difference in destination-choice pattern between black and white elderly
primary migrants can be vividly depicted by the migration flows from the southern
states. Table 1 shows the three most attractive destinations, together with their percen­
tage shares, of the 'race'-specific elderly primary migrants from each of the southern
states in 1985 - 90. For black people, many of these destinations are the industrial states
Table 1. The three most preferred destinations of (a) black and (b) white elderly primary migrants
from the southern states in the period 1985-90 [source: 1990 Census public use microsample
(PUMS)].
Origin
state
(a)
Virginia
West Virginia
North Carolina
South Carolina
Georgia
Florida
Kentucky
Tennessee
Alabama
Mississippi
Arkansas
Louisiana
Oklahoma
Texas
Best
destination
Maryland
Virginia
Virginia
New York
Florida
Georgia
Indiana
Michigan
Florida
Illinois
California
Texas
California
California
Second best
Third best
share destination
share destination
share
(%)
(%)
(%)
29.2
19.1
19.3
16.4
31.2
17.6
48.1
26.9
16.3
22.8
20.3
27.7
42.0
49.5
New York
California
New York
Pennsylvania
Ohio
New York
California
Illinois
Georgia
Tennessee
Illinois
California
Texas
Colorado
17.7
12.5
17.0
13.6
12.2
12.1
13.1
17.2
12.3
14.5
14.9
27.0
25.2
9.4
New Jersey
Kentucky
Maryland
North Carolina
New York
Maryland
Illinois
Indiana
Ohio
Louisiana
Michigan
Ohio
Michigan
Oklahoma
13.6
12.5
11.7
12.9
7.9
8.5
11.4
9.1
12.2
9.2
14.3
5.5
7.4
8.1
No. of
migrants
1312
168
1434
1639
1422
637
511
618
1881
2180
847
1536
421
1308
(b)
Virginia
29.3 North Carolina 21.4 Maryland
8.1
7272
Florida
West Virginia Florida
28.3 Ohio
21.5 Virginia
9.7
6736
North Carolina South Carolina 25.5 Florida
22.1 Virginia
13.9
5988
South Carolina North Carolina 31.2 Florida
21.6 Georgia
19.1
2183
39.0 Alabama
Georgia
Florida
12.3 South Carolina 10.3
6098
Georgia
Florida
30.0 North Carolina 16.2 Alabama
9.1
3005
25.5 Ohio
19.9 Indiana
Kentucky
Florida
17.6
9838
18.7 Georgia
Tennessee
Florida
11.6 Mississippi
11.1
6955
21.4 Tennessee
Alabama
Florida
33.3 Georgia
11.6
6450
12.4
3184
Mississippi
Tennessee
20.8 Louisiana
14.3 Alabama
Arkansas
Texas
22.0 California
10.6 Oklahoma
10.6
4577
30.3 Mississippi
23.0 Florida
7.9
4622
Texas
Louisiana
13.6 Arkansas
6505
Texas
Oklahoma
28.9 California
8.1
California
14.4 Arkansas
Texas
12.0 Oklahoma
ILl 11702
Note; Delaware, Maryland, and Washington, DC, are included in the northern industrial
region in this study.
, .
.
Adult children and elderly migration
201
in the snowbelt (for example, Illinois, New York, Michigan, Pennsylvania, Ohio, and
Indiana). For white people, few of these destinations are in the snowbelt.
Although black migrants' preferences for northern industrial states have been
frequently attributed to the relatively generous welfare and social programs of these
states,(9) it is more plausible to expect that the elderly black migrants were attracted
mainly by their adult children, many of whom happened to be located in the industrial
north. To illustrate (but not to prove) the plausibility of this expectation, we have
plotted the destination choice patterns of the widowed and married elderly primary
black migrants from Alabama against the distribution of their adult children (figure 1).
Figure 1(a) shows that the northern industrial states that attracted large proportions of
widowed black migrants indeed tended to have large shares of their adult children. The
relationship between the distribution of adult children and the destination choice
pattern of the widowed elderly appears to be rather strong (R2 = 0.729). Figure l(b)
shows that the corresponding relationship is much weaker for married elderly black
people (R 2 = 0.341). The difference for widowed and married elderly black people is
consistent with Litwak's characterization of the MEF in the sense that the elderly who
are less capable of living independently are more likely to move toward their adult
children.
Regarding the attraction of elderly white people to their adult children, figures I (c)
and l(d) show the relevant information for widowed and married white primary
migrants from Alabama. The data suggest that the attraction to adult children was
very strong and hardly differed for the widowed and the married (R2
0.912 and
0.910, respectively). Widowed and married elderly white people were strongly attracted
16
~
..,
"""il
'­
0!l
.., C
00«1
«I ....
... 00
u.e
.......
... :::: .,;'­
OJ
~o
R2 = 0.729
12
10
8
6
~
, FL
CA
~
IN
4
NY
i
W'
;~
R2
0.912
25
' GA
20
5
TN
0
MI
Itr!i:"~!" "
, TX
k
0"""
(c)
(a)
35
>.
FL
N = 2699
30
5
2
0
35
,. IL
GA
N = 1063
14
30
35 FL
N = 322 25
25
'­
20
20
°~ E
'" 15
!loo
~.~ 10
.......
.., :::: ~o
5
!
PA
o~.t~1~»:
o
NJ
" IN
MI
ILy
, GA
15
TN
, ­ 10
.,CA
hMI
"OH
5
GAY
7
8
2
3
4
5
6
Percentage of Alabama-born black 'adult
children' who resided in other states in 1985
(b)
FL
N = 2877
R2 = 0.910
30
R2 = 0.341
.:::OJ
~
o.~::-.
0
"
.
TX
w
4
2
3
6
7
8
5
Percentage of Alabama-born white 'adult
children' who resided in other states in 1985
(d)
Figure 1. Attraction of Alabama's 1985 - 90 black and white elderly primary out-migrants to their
adult children who resided in other states in 1985, by the elderly's marital status: (a) black
widowed, (b) black married, (c) white widowed, (d) white married.
For a list of references about the potential effects of the spatial variation in welfare and social
benefits on black migrants (see Long, 1988, page 149).
(9)
\
'
K-L Liaw, W H Frey, J-P Lin
202
to Florida and Georgia, which had not only a large concentration of their adult
children but also an attractive environment. Without multivariate analysis it is difficult
to know the differential attractions of these two factors. The main message from these
figures is that the attraction to adult children should not be forgotten when attempting
to explain the flows of elderly migrants into amenity-rich states. (10)
We now turn to the estimation results of the destination choice submodel. In the
best specification we find that the elderly primary migrants of both 'races' and for all
marital statuses were attracted to their adult cliildren,ol) that this attraction was some­
what stronger for the relatively old (aged 75 years and over) widowed, and that it was
substantially stronger for white people than for black people (table 2). To be more
specific, we can compute the estimated odds ratios for four separate groups of elderly
migrants:
(1) For white people younger than 75 years and not widowed, the odds ratio is
exp (0.291) = 1.34, which means that if the share of adult children by a potential desti­
nation is increased by one percentage point, the odds that this potential destination
is selected will be increased by a factor of l.34.
(2) For widowed white people aged 75 years and over, the odds ratio is
exp (0.291 + 0.020) = 1.36.
(3) For black people younger than 75 years and not widowed, it is
exp(0.291 - 0.151) = 1.15.
(4) For widowed black people aged 75 years and over, it is
exp(0.291 + 0.020 - 0.151) = 1.17.
The finding that elderly black people were substantially less attracted to their out­
migrated adult children than were elderly white people is consistent with the black­
white difference demonstrated by figure 1. It is also consistent with the finding of
Hogan et al (1993) that black families tend to have weaker intergenerational connec­
tions than do white families. It suggests that black families tend to differ more from the
model of the MEF than do white families. The finding that the widowed were more
likely to be attracted to adult children than were those of other marital status is
consistent with the greater attraction of California for widowed compared with mar­
ried elderly migrants, as mentioned earlier.
The estimated coefficients of the amenity variables show that the elderly migrants
in general tended to be attracted to potential destinations with a warmer winter, clearer
days, and a location on the Gold Coast. The group that was most subject to the
attraction of a warm winter was that of nonpoor married white people in the 65 - 69
year age interval.(12) Compared with their white/. counterparts, all cohorts of black
FI".;·IV
(10) In two small-scale surveys of the post-retirqfuent migrants from north central states to the
amenity-rich states of Arizona (N = 199) and 'N = 150), it was found that 54% "had at least
one child located closer now than before their retirement", and that "the desire to be nearer
children was, in fact, indicated by 31 percent of the Arizona migrants, and by 13 percent of those
in Florida, as a primary consideration in their decision to retire outside their home communities"
(Bultena and Marshall, 1970, page 91).
(11) The never-married elderly in general did not have a child and represented a very small propor­
tion of the elderly popUlation, having little effect on the estimated results. Our interpretations of
the statistical results about the attraction of children do not apply to this group.
(12) The age pattern of the odds ratio of the coldness variable for unmarried and nonpoor white
migrants is: exp( - 0.28) = 0.76 for the 60-64 age-group, exp( - 0.28 - 0.11) = 0.68 for the 65­
69 age-group, exp ( - 0.28 - 0.08) = 0.70 for the 70-74 age-group, exp ( - 0.28) = 0.76 for the
75 - 79 age-group, and exp ( - 0.28 + 0.07) = 0.81 for the 80 plus age-group. These odds ratios,
together with other odds ratios that can be computed in a similar way, indicate that white migrants
of every age-group tended to avoid destinations with a relatively cold winter, and that white
migrants of the 65 - 69 age-group had the strongest aversion to a cold winter (or, equivalently the
strongest attraction to a warm winter). For black migrants who were married, aged 65 - 69 years,
\
.
Adult children and elderly migration
203
Table 2. Estimation results of the destination choice submodel for the 1985 - 90 interstate black
and white elderly (aged 60 years or more in 1990) primary migrants in the United States.
Best specificationa
Explanatory variable
Attraction to adult children
(adult children)
(adult children) x (widowed) x (aged 75 or more yeats)
(adult children) x black
Environmental amenity:
climate
(coldness of winter)
(coldness of winter) x married
(coldness of winter) x (aged 65-69 years)
(coldness of winter) x (aged 70-74 years)
(coldness of winter) x (aged 80 or more years)
(coldness of winter) x poor
(coldness of winter) x black
cloudiness
scenic beauty and recreational opportunity
Gold Coast
(Gold Coast) x black
Relative location
In(distance)
In(distance) x married
In(distance) x (post-secondary education)
In(distance) x Alaska
In(distance) x Hawaii
contiguity
contiguity x black
Cost of living
(cost-of-living index)
(cost-of-living index) x black
Generosity of medical programs
Medicare
Medicaid
Medicaid x black
Racial attraction
(racial similarity)
(racial similarity) x black
Labor-market variables
income x (aged 60-64 years)
(employment growth) x (aged 60-64 years)
Size of ecumene
In(population size)
In(population size) x black
p 2 -statistic
coefficient
t-ratio
0.291
0.020
-0.151
62.7
2.8
-15.3
MCR
0.0264
0.0563
0.0298
-0.280
-0.186
-0.113
-0.078
0.066
0.040
0.449
-0.103
-22.0
-21.1
-12.8
-7.9
6.4
3.2
15.8
-24.8
0.989
-0.561
35.3
-5.4
-0.566
-0.117
0.043
0.349
0.075
0.710
-0.244
-27.6
-5.8
2.6
12.4
4.3
22.7
-2.9
-0.074
0.072
-31.8
10.6
0.334
0.153
-0.179
9.4
27.1
-8.7
0.532
-0.191
31.4
-4.0
0.666
4.188
5.2
11.6
0.174
0.420
11.1
8.2
0.0070
0.0305
0.0058
0.0042
0.0057
0.0015
0.0015
0.3507
Note: MCR, marginal contribution to the p2- statistic; for more details on the explanatory
variables, see text, sections 3 and 4 and the appendix.
a See text, section 3.2.
continued
and nonpoor, the odds ratio of coldness is exp ( - 0.28 - 0.19 - 0.11 + 0.45) = 0.88, compared
with the very low value of exp( - 0.28 - 0.19 - 0.11) = 0.56 for their white counterparts. In other
words, the black migrants had a much weaker aversion to destinations with a cold winter than did
their white counterparts.
(12)
\
.
204
K-L Liaw, W H Frey, J-P Lin
people were much less attracted by destinations with a warm winter. Although both
black and white elderly migrants were subject to the attraction of the states on the
Gold Coast, the attractions were much weaker for black people than for white people.
From the marginal contributions in the p 2-statistic (table 2), we see that, although
the explanatory power of the adult children variable was much less than the combined
explanatory power of the variables climate and Gold Coast, it was stronger than that
of Gold Coast. Compared with the covariates that are not the focus of this paper, adult
children was less powerful than the combination of distance(13) and contiguity but
much more influential than cost of living, generosity of medical programs, (14) racial
similarity, labor-market variables (in the 60 - 64 age-group), and population size at
destination. Overall, we find that the adult children variable was one of the most
important explanatory factors in the destination-choice submodel.
Table 3. Observed and predicted shares of elderly primary migrants (aged 60 or more years in
1990) in 1985 - 90 by the major receiving states: (a) black people and (b) white people.
Rank Destination
(a)
I
2
3
4
5
6
7
8
9
10
In-migrants a
Share (%)
Native
pop. in
observed predicted observed predicted 1985a
California
Maryland
Florida
New York
Illinois
Georgia
Michigan
New Jersey
Virginia
Texas
In-migration rate (%)
observed predicted
2578
1749
1713
1394
1215
I 120
1104
1054
1001
958
2860
1593
1448
1881
I 711
I 187
751
845
745
871
11.2
7.6
7.5
6.1
5.3
4.9
4.8
4.6
4.4
4.2
12.5
6.9
6.3
8.2
7.4
5.2
3.3
3.7
3.2
3.8
15328
48468
72816
59925
34291
151891
19907
24778
108049
172117
16.8
3.6
2.4
2.3
3.5
0.7
5.5
4.3
0.9
0.6
18.7
3.3
2.0
3.1
5.0
0.8
3.8
3.4
0.7
0.5
13886
22967
13892
22967
60.5
100.0
60.5
100.0
707570
1663973
2.0
1.4
2.0
1.4
Florida
211481
Arizona
31871
California
31605
20589
Texas
New Jersey
19687
North Carolina 16322
Pennsylvania
12591
Virginia
11508
Georgia
10496
South Carolina 10471
190024
28654
41458
21371
11769
14870
12932
14006
22507
IS 311
37.7
5.7
5.6
3.7
3.5
2.9
2.2
2.0
1.9
1.9
33.8
5.1
7.4
3.8
2.1
2.6
2.3
2.5
4.0
2.7
190933 110.8
24475 130.2
711631
4.4
1090723
1.9
614173
3.2
634209
2.6
1888668
0.7
418173
2.8
463153
2.3
257339
4.1
99.5
117.1
5.8
2.0
1.9
2.3
0.7
3.3
4.9
5.9
Top-10 destinations 376621
All destinations
561435
372 902
561435
67.1
100.0
66.4
100.0
Top-10 destinations
All destinations
(b)
I
2
3
4
5
6
7
8
9
10
6293477
20170529
6.0
2.8
5.9
2.8
Number of persons.
Note: pop., popUlation.
a
The distance-decay effect is flatter for the far-away states of Alaska and Hawaii.
Although both black and white migrants were attracted by the destinations with a more
generous Medicare program, only white migrants were significantly attracted by destinations with
a more generous Medicaid program. In other words, our result does not support the idea that
elderly black people were more prone to be attracted by states with more generous government
programs than were elderly white people.
(13)
(14)
I.
•
Adult children and elderly migration
205
Although the p 2 -statistic (0.3507) of the best specification in table 2 may appear
small to a reader who is familiar with the R 2 -statistic as a measure of goodness of fit,
we see in table 3 that the destination-choice submodel actually predicts quite well the
observed patterns for black people and white people. It is interesting to note that two
of the top three destinations of white elderly primary migrants, namely Florida and
California, had the largest shares of nonnative white adult children among all the
states in 1985: 8.7% in Florida, and l3.7% in California, compared with only 3.1 % in
New York. An important black - white contrast in· destination-choice pattern is that
Florida attracted by far the largest number of elderly white primary migrants (211481),
but it attracted significantly fewer elderly black primary migrants (17l3) than did
California (2578), even though it is much closer to the major concentration of elderly
native black people than is California. This difference is consistent with the fact that
Florida had only 5.3% of the nonnative black adult children in 1985, compared with
14.0% for California. As New York had as many as 10.4% of nonnative black adult
children in 1985, it is not surprising that it was among the top five destinations of
elderly black primary migrants.
S.2 Departure choices
As expected, the estimated coefficients of the 'adult children' variable and its inter­
action term in the departure submodel show that the concentration of adult children in
their state of birth had a significant retention effect on both native elderly black people
and white people and that this effect was stronger for the widowed than for those of
other marital statuses (table 4, see over). The coefficients of environmental factors also
turn out to be consistent with our expectations: the states on the Gold Coast were in
general more capable of retaining their elderly natives, whereas states with a colder
winter were more prone to 'push out' those elderly natives who were married couples
or males at retirement age. The marginal contributions in the p2- statistic show that,
among the factors representing place attributes, adult children and environmental
amenities, together with cost of living, were most important: they were more important
than generosity in medical programs, proportion of homeownership, rate of violent
crime, population size at origin, and location in the industrial heartland. (15)
The marginal contributions in the p 2 -statistic also show the typical finding that
personal attributes were in general more important than place attributes in determin­
ing departure propensities. The large negative coefficient of the variable black (-0.946)
indicates that native elderly black people were much less likely to migrate than were
native elderly white people. This difference is also reflected by the observed departure
rates: 1.4% for black people compared with 2.8% for white people. By contrast, the
small positive coefficient of the variable male (0.036) indicates that the migration
propensities did not differ much by gender. This is reflected by the small gender
difference in the observed departure rates: 2.6% for females compared with 2.8% for
males. The estimated coefficients also show that the retirement peak of the departure
(out-migration) schedule was quite clear for white people but hardly existed for black
people. The observed age-specific departure rates also display a retirement peak for
white people: 3.2% for the 60-64 age-group, 3.3% for the 65-69 age-group, and 2.6%
In the multivariate context, we found that the industrial heartlands variable had a significant
retention effect on native elderly black people and a significant push effect on native elderly white
people. This finding suggests that most of the native elderly black people were much less affiuent
than the corresponding white people in the industrial heartland so that they were less prone to
participate in amenity-oriented migration. The changes in the coefficients of other variables arising
from the deletion of this dummy variable (not shown in this paper) suggest that the black - white
difference in the propensity to leave the industrial heartland was partly a result of a high proportion
of black adult children born in this region remaining within the region.
(15)
, .
206
K-L Liaw, W H Frey, J-P Lin
Table 4. Estimation result of the departure submodel for the interstate migrations by the black
and white elderly (aged 60 years or more in 1990) native-born Americans in 1985 - 90.
Best specificationa
Explanatory variable
coefficient
t-ratio
Constant term
-4.859
-18.9
Race
black
-0.946
-9.5
Gender
0.036
male
2.1
Age
[retirement age (65 years)]
0.159
6.8
[retirement age (65 years)] x black
-0.135
-1.7
Marital status
-0.167
single
-3.1
single x black
1.311
9.8
(divorced or separated)
0.747
16.6
(divorced or separated) x black
0.604
5.9
widowed
1.306
9.5
widowed x black
0.620
6.9
widowed xfemale x (aged 75-79 years)
0.294
6.7
widowed x female x (aged 80 or more years)
15.1
0.564
widowed x male x (aged 75-79 years)
0.345
4.4
widowed x male x (aged 80 or more years)
7.7
0.479
Educational attainment
15.5
(secondary education)
0.258
(post-secondary education)
34.8
0.600
(post-secondary education) x blacks
-0.307
-3.4
Poverty status
-0.211
poor
-7.6
poor x black
-0.515
-6.0
Retention effect of adult children
(adult chlldren)
-0.016
-10.5
(adult chlldren) x widowed
-0.016
-7.9
Environmental amenity:
climate
coldness x married
14.8
0.106
coldness x (retirement-aged male)
0.017
3.3
scenic beauty and recreational opportunity
(Gold Coast)
-8.1
-0.227
Cost of living
20.1
(cost-of-living index)
0.030
Generosity in Medical Programs
Medicare x (aged 75 or more years)
-0.025
-2.1
Medicaid x (aged 75 or more years)
-0.035
-6.4
Homeownershlp
(homeownershlp proportion)
-11.0
-0.020
Social environment
(violent crime rate)
3.253
7.7
Size of ecumene
In(population size)
0.039
2.3
Regional effect
11.9
(industrial heartland)
0.249
(industrial heartland) x black
-0.471
-5A
Attraction of rest of system
inclusive variable
7.9
0.076
P2-statistic
0.0424
Ii See text, section 3.2.
Note: MCR, marginal contribution to the p2-statistic; for more details on
variables, see sections 3 and 4 and the appendix.
\
.
MCR
0.0014
0.0013
0.0025
0.0057
0.0054
0.0007
0.0014
0.0015
0.0011
0.0003
0.0017
0.0008
0.0005
0.0003
0.0000
0.0007
0.0003
the explanatory
Adult children and elderly migration
207
for the 70 - 74 age-group. The corresponding rates for black people are 1.3%, 1.2%, and
1.3%, respectively.
To the extent that educational attainment and poverty status serve well as proxies
for socioeconomic status, the estimation result confirms that the higher the socio­
economic status, the greater the migration propensity. The estimated coefficients
indicate that educational attainment had highly significant positive effects, which
were stronger for white people than for black people. The observed departure rates
are consistent with this multivariate finding: the rates for white people were 2.0% (less
than high school), 2.9% (high school), and 4.0% (college), whereas the corresponding
rates for black people were 1.2%, 1.6%, and 1.9%, respectively. With respect to the
effects of poverty status, the estimated coefficients and the observed departure rates
show that those under the poverty line were less likely to migrate. However, without
controlling for other factors, the observed departure rates were unable to substantiate
the multivariate finding that the negative effect of poverty was stronger for black
people than for white people. (16)
The estimated coefficients of the dummy variables representing marital status and
their interactions with race, age, and gender suggest that the effects of marital status
were relatively complex. With the minor exception of never-married white people, the un­
married were more migratory than the married. Among the unmarried, the widowed
were most migratory. The contrast between the widowed and the married was much
greater for black people than for white people. For each gender, the propensities of the
widowed to undertake primary migration increased monotonically with age beyond
the early seventies. It is important to realize that these findings are obtained from a
multivariate framework whereby other explanatory variables have already represented
(I) the greater tendency of the widowed to be retained by states containing many of
their adult children, and (2) the greater tendency of the married to be pushed out
of states with a relatively cold winter. Mainly because of these two tendencies, the
observed departure rates turned out to be somewhat lower for the widowed (2.4%) than
for the married (2.8%).
To follow the main theme of this paper, we end this section by focusing on the
marked difference between the overall departure rates of black people and white
people (1.4% compared with 2.8%), which are perfectly predicted by the departure
submodel. This difference can be partly accounted for by the fact that elderly black
people had a lower educational attainment, were more likely to be in poverty, and
were less prone to migrate at retirement age. It can also be related to the fact that a
high proportion of black adult children had made a lifetime migration from the
sunbelt to the snowbelt, whereas the opposite was true for white adult children. In
other words, the difference in departure rates was also attributable to the fact that
the attraction to adult children and environmental amenities countered each other
for elderly black people but reinforced each other for elderly white people.
5.3 Interstate net transfers of elderly primary migrants
Although the elderly are much less migratory than are young adults, the elderly
migration process tends to operate "like a giant parabolic mirror, collecting distinctive
types of individuals from everywhere and concentrating them into certain places"
(Morrison, 1990, page 401). It can thus result in fairly large net gains for a few states.
Realizing the possibility of attracting a sizable number of elderly migrants to their
territories, some southern states and municipalities have developed strategies and
(16) The observed departure rates for white people are 1.8% (poverty) and 2.9% (non poverty). The
corresponding rates for black people are 0.9% (poverty) and 1.5% (non poverty).
\
.
K-L Liaw, W H Frey. J-P Lin
208
Table 5. (a) The top-ten net gainers and (b) top-ten net losers of black and white elderly primary
migrants in 1985 - 90: observed and predicted patterns.
Rank Top-10
(a)
Black
1
2
3
4
5
6
7
8
9
IO
people
California
Maryland
Florida
Michigan
New Jersey
Indiana
Ohio
Illinois
Wisconsin
Nevada
Total
White people
Florida
I
Arizona
2
California
3
4
North Carolina
Nevada
5
Texas
6
7
South Carolina
Oregon
8
Georgia
9
Arkansas
10
Total
Net migrantsa
observed
predicted
2363
1411
1076
817
593
564
538
466
423
289
8540
2645
1017
844
525
397
171
300
1214
226
117
7456
208476
31281
14609
10334
9587
8887
8288
5487
4398
4309
305656
Native
population
3
in 1985
Net migration rate (%)
observed
predicted
15328
48468
72816
19907
24778
12426
36275
34291
878
41
265208
15.4
2.9
1.5
4.1
2.4
4.5
1.5
1.4
48.2
704.9
3.2
17.3
2.1
1.2
2.6
1.6
1.4
0.8
3.5
25.7
285.4
2.8
187470
28016
24645
7820
6792
7848
12721
2731
16681
1408
296132
190933
24475
711631
634209
7351
1090723
257339
131205
463153
249111
3760130
109.2
127.8
2.1
1.6
130.4
0.8
3.2
4.2
0.9
1.7
8.1
98.2
114.5
3.5
1.2
92.4
0.7
4.9
2.1
3.6
0.6
7.9
-2011
-1603
-1280
-1232
-698
-680
-474
-393
-350
-311
-9032
-1440
-1627
-708
-1098
-635
-897
449
-143
-601
-605
-7305
107775
130 lSI
114759
131543
42507
148201
59925
27002
172117
108049
1042029
-1.9
-1.2
-1.1
-0.9
-1.6
-0.5
-0.8
-1.5
-0.2
-0.3
-0.9
-1.3
-1.3
-0.6
-0.8
-1.5
-0.6
0.7
-0.5
-0.3
-0.6
-0.7
-112337
-38724
-32200
-24595
-23101
-20921
12055
-8597
-8547
-7955
-289032
-96340
-44639
-36517
-24588
-23156
-25627
-15007
-4152
-6718
-13438
-290182
1930530
1189369
1888668
754614
859853
I 155149
614173
578840
428723
633555
10033474
-5.8
-3.3
-1.7
-3.3
-2.7
-1.8
-2.0
1.5
-2.0
-1.3
-2.9
-5.0
-3.8
-1.9
-3.3
-2.7
-2.2
-2.4
-0.7
-1.6
-2.1
-2.9
(b)
Black people
Mississippi
I
Alabama
2
South Carolina
3
4
Louisiana
Arkansas
5
North Carolina
6
New York
7
Kentucky
8
Texas
9
Virginia
10
Total
White people
I
New York
Illinois
2
Pennsylvania
3
4
Massachusetts
Michigan
5
Ohio
6
New Jersey
7
Indiana
8
Iowa
9
Wisconsin
10
Total
a Number of persons.
\
.
209
Adult children and elderly migration
programs to attract relatively well-off elderly migrants as a way to boost "population,
incomes, and employment" (Glasgow and Reeder, 1990, page 434).
Table 5 shows that the major net gains and net losses in elderly primary migration
are fairly well predicted by the explanatory factors used in the nested logit model.
More importantly, they also show that the interstate net transfers of elderly primary
migrants were voluminous for white people and rather small for black people, and that
the net transfers of white people were strongly oriented from amenity-poor states to
amenity-rich states, whereas the net transfers of black people were mainly from south­
ern states to a mixture of northern industrial states and a few amenity-rich states.
These differences reflect the important fact that the adult children who had made the
birth-to-1985 interstate migration were distributed quite differently in 1985 for these
two cohorts: 67% of white people and 52% of black people settled down in South and
West regions where amenity-rich states are concentrated. The white adult children's
greater concentration in amenity-rich states helped to pull a greater proportion of their
parents to the sunbelt, whereas the relatively high concentration of black adult children
in northern industrial states generated a northward pull that largely canceled out the
southward pull by environmental amenities. This finding suggests that the amenity-rich
states with a higher concentration of nonnative adult children are more likely to
succeed in their attempt at attracting relatively well-off elderly primary migrants for
boosting their economic base.
6 Explanation for the unexpected finding of Longino and Serow
In a paper on the characteristics of elderly return migrants, Longino and Serow
formulated the hypothesis that "although there will be regional variation, return
migrants are more likely to be older and more widowed and residentially dependent
than nonreturn migrants for the nation and for all regional streams" (1992, page S39).
They tested this hypothesis with the census data on the 1975 - 80 interstate elderly (aged
60 or more years in 1980) in-migrants of the four census regions of the United States.
The hypothesis was well supported at the national level but was significantly contra­
dicted by the data of the Midwest. At the national level, among the return migrants
24.5% were aged 75 or more years, 33.0% were widowed, and 75.6% lived independently,
whereas among the nonreturn migrants, the corresponding figures were 21.5%, 27.0%,
and 79.5%, respectively. However, in the Midwest, these figures were 28.6%, 34.9%, and
76.7%, respectively, for the return in-migrants, compared with 32.2%, 41.4%, and 65.8%,
respectively, for the nonreturn in-migrants.
Why was the hypothesis contradicted so sharply by the elderly in-migrants of the
Midwest? Longino and Serow did not attempt to explain this contradiction specifically
but speculated in the concluding section that, for the elderly, "one's informal support
system, composed of close friends and children, may more often be located at one's
adult state of residence than at one's state of birth" (1992, page S42). We think that the
answer lies partly in this speculation in the sense that the attraction to adult children
was recognized as a potentially important factor. But a plausible answer also requires
the recognition that many adult children may have migrated to a state that is neither
their state of birth nor the state of previous long-term residence of the elderly. As the
results of our multivariate analysis indicate that the location of adult children and
environmental amenity are among the most important place attributes for accounting
for elderly primary migration, and as the Midwest is in general perceived to be lacking
in environmental amenity, it is likely that many elderly in-migrants to the Midwest are
attracted mainly by their previously out-migrated adult children, because of their (the
elderly's) need for assistance, and are hence more likely to be older and widowed and
to coreside with their adult children. We now use 1990 Census data to see if some
\
.
K-L Liaw, W H Frey, J-P Lin
210
Table 6. Different characteristics of nonreturn (primary plus onward) and return interstate
elderly in-migrants to the US Midwest: 1985 - 90.
Characteristic
Return
migrants
Nonreturn migrants
primary
onward
total
62775
25
40
41
91556
37
33
34
154331
62
36
37
96286
38
32
34
(b) Black people
Number of in-migrants
Sharea
Number aged 75 or more yearsb
Number widowed"
5237
33
43
52
8616
55
34
46
13853
88
37
48
1842
12
21
39
(c) White people
Number of in-migrants
Sharea
Number aged 75 or more yearsb
Number widowedc
57538
24
40
40
82940
35
33
33
140478
60
36
36
94444
40
33
34
(a) Black people and white people
Number of in-migrants
Sharea
Number aged 75 or more yearsb
Number widowed c
,~
As a percentage of the number of US-born elderly in-migrants of the US Midwest.
b As a percentage of all people aged 60 or more years in the given cohort.
C As a percentage of people of all marital statuses in the given cohort.
a
Table 7. Different characteristics of black and white elderly primary interstate in-migrants of the
Midwest from different regions: 1985 - 90.
Region of origin
Characteristic
Northeast
(a) Black people
Number of in-migrants
Sharea
Number aged 75 or more years b
Number widowedc
Midwest shared
(b) White people
Number of in-migrants
Sharea
Number aged 75 or more yearsb
Number widowed"
West
total
212
4.0
46
37
750
14.3
20
34
4251
81.2
46
56
24
0.5
6
34
25
8
23
7451
36189
12008
1890
57538
12.9
46
47
Midwest shared
Midwest South
3
62.9
38
36
20.9
43
47
20
13
5237
100
43
52
3.3
35
33
100
40
40
5
10
Note: as the actual sample size of black migrants from the states in the West region is only
about 2 persons (24 x 0.08), the values of their characteristics are not shown.
a As a percentage of the total.
b As a percentage of all aged 60 or more years in the given cohort.
" As a percentage of people of all marital statuses in the given cohort.
d Of elderly primary migrants from the states of each region, for the given cohort.
\
.
Adult children and elderly migration
211
suggestive evidence for this explanation can be found (tables 6 and 7). For simplicity,
we now refer to the older and more likely to be widowed elderly as the 'needy' elderly.
In table 6 we divide the 1985-90 US-born black and white elderly interstate in­
migrants of the Midwest into return and nonreturn cohorts, the latter being further
divided into primary and onward cohorts. In table 6(a) we can see that nonreturn
migrants are indeed needier than are return migrants and that the primary component
is the 'neediest'. In tables 6(b) and 6(c) we can see that these statements are true for
black people and white people. Thus, our focus is on elderly primary migrants.
In table 7, we divide the black and white primary in-migrants of the Midwest into
those from the states of each of the four census regions. In table 7(a) we can see that a
large majority of the black primary in-migrants (81.2%) came from southern states and
that the black primary migrants from southern states are the neediest. Because as
many as 82.2% of the nonnative black adult children in the Midwest in 1985 were
born in the South, it is highly plausible to claim that the black needy elderly primary
(and hence nonreturn) in-migrants of the Midwest were attracted mainly to their adult
children. In table 7(b) we can see that the neediness of the white elderly in-migrants of
the Midwest is attributable to the neediness of the migrants coming from the South
and Northeast, who represented only a small fraction (13% and 3%, respectively) of the
white elderly primary out-migrants from the states of these two regions. Instead of
joining the dominant streams towards the amenity-rich states in the South, these needy
primary migrants came to the Midwest where the winter is cold. Although the Midwest
does not have a high proportion of white adult children born in the South and North­
east, our multivariate results suggest that the needy elderly white people in the small
tributaries flowing from the South and Northeast into the Midwest were attracted
mainly by their adult children. These small tributaries undoubtedly included the elderly
primary migrants from the economically depressed areas of West Virginia and western
Pennsylvania who went to join the households of their adult children in the industrial
cities of the Midwest (Rowles, 1983). In sum, our evidence suggests that the unexpected
finding of Longino and Serow (1992) is the consequence of the attraction of elderly
primary migrants to their adult children.
Longino and Serow (1992) have highlighted that the general characteristics of a
given type of migrant observed in different regions may be quite different from those
observed at the national level and that this difference can help to yield greater insight
into the major determinants of elderly migration. For example, the fact that the return
in-migrants of the West and South tend to be significantly younger and less likely to be
widowed than return in-migrants of the Northeast and Midwest was used by them to
infer the importance of what Cribier (1980) described as 'provincial return migration'.
In other words, an important motivation for an older person's return migration after
his or her retirement from a modestly successful working career is the lower cost of
living and the higher social status that he or she can enjoy in the region of his or her
family roots. Similarly, the fact that the elderly primary in-migrants of the Midwest are
unusually old and often widowed can be used, in our opinion, to suggest the impor­
tance of the attraction to adult children who are willing to assist their weakened elderly
parents.
7 Concluding discussion
We have presented our study of black and white elderly primary migration in the
context of Litwak's theory that the only family system that is consistent with an
industrialized and bureaucratized society is the system of modified extended family,
which legitimizes the out-migration (as well as social mooility) of adult children for
career advancement and encourages the migration of elderly parents to be close to
\
.
212
K-L Liaw, W H Frey, J-P Lin
their adult children for the assistance that requires proximity. In our opllllon, this
family system is much better than the alternative system of the isolated nuclear family
or the highly individualized mass society, especially as aging of the population has
become the dominant demographic trend of the new century.
By using a reasonable proxy for the location of adult children we have shown that
married and unmarried elderly natives were strongly attracted to their adult children,
although the attraction was stronger for the widowed. In the departure process we
found that the elderly (especially those who were widowed) were more likely to remain
in the states where a higher proportion of their adult children had remained. In the
destination choice process, we found that elderly migrants were more likely to move to
states where a higher proportion of their migrated adult children were located and
that this tendency was somewhat stronger for those who were widowed and aged 75
years and over. These findings can be taken as hopeful signs of the viability of the
MEF system. They also suggest that the elderly natives did not have a strong tendency
to delay their migration towards their adult children until the loss of a spouse or
becoming very old.
We have also shown that the attraction of adult children and environmental ameni­
ties were weaker for elderly black people than for elderly white people. This finding is
related to the fact that a relatively high proportion of black adult children had made
lifetime migration from the sun belt to the industrial states of the snowbelt, whereas a
relatively high proportion of white adult children had made life-time migration in the
opposite direction. Thus, the two types of attractions were more likely to counter each
other for native elderly black people and to reinforce each other for native elderly
white people. (17) As a consequence, the migration of native elderly black people was
very small in volume and was somewhat oriented towards the snowbelt, whereas the
migration of native elderly white people was rather large in volume and was strongly
oriented towards the sunbeIt.
In light of the highly controversial literature on black families in the United States
(for a review, see Franklin, 1997) we hesitate to make a strong connection between our
finding about the weaker attraction of the black elderly to their adult children and the
finding of Hogan et al (l993) that black people are consistently less likely than are
white people to be involved in intergenerational assistance.(lS) This connection would
(17) Note that the countervailing and reinforcing effects of the location of adult children and
environmental amenities appeared in an additive fonn in our model. One of the referees mentioned
a scenario: a school teacher in midcareer might select a promotion in, and move to, an amenity­
rich region (and away from her parents) knowing that 5 15 years later her parent(s) will retire and
that they are keen to move to an amenity-rich area and to live near her. The question about this
scenario is: can the amenity-family motivations be easily distinguished in either generation?
The answer is no. But, if there were many cases like this, there would be significant interaction
(multiplicative) terms between the proxies of these two factors in the nested logit model.
(18) Based on a three-level nested logit model, our multivariate analysis of black and white elderly
return and onward migration behavior has revealed the following main points. First, elderly non·
natives were strongly attracted to the location of their adult children when they made their
migration decisions, at all levels of the choice framework (to depart or to stay put; to return or
to move onward; and to select one of the many potential destinations for onward migration).
Second, at all three levels, the attraction to the location of adult children was stronger for the
widowed than for those of other marital statuses. Third, in the return/onward-choice and destina­
tion-choice submodels the attraction to adult children was much weaker for black people than for
white people. This finding suggests that intergenerational connections were weaker for black
people than for white people, despite the fact that since the 1960s elderly black people have been
more likely to coreside with their kin (including their adult children) than have elderly white people
(Ruggles, 1994). In short, the main findings from our analysis of primary migration are reinforced
by the findings from our analysis of return and onward migration. The results of our analysis of
return and onward migration are available to the reader on request.
, .
"
Adult children and elderly migration
213
suggest that the structure of many black families has been less conducive to the
development of long-term commitments to family members. One would suggest that
perhaps the emphasis on commitments by the Million Man March is not as pointless
as some cynics may suggest.(19)
Owing to the lack of data on the specific locations of noncoresident adult children,
the importance of the attraction to adult children has not been assessed by other
researchers in the context of environmental amenities and other relevant factors.
Thus, the empirical implications of this attraction have remained largely unexplored.
In this paper we have demonstrated that it can help to explain the 'unexpected' contrast
between return and nonreturn elderly in-migrants of the Midwest: the latter are older,
more likely to be widowed, and less likely to live independently. We hope that our
findings will provide a basis for encouraging collection of information on the specific
locations of non-coresident adult children in censuses and surveys. (20)
In light of the trend of the extremely rapid growth of the oldest old in the United
States (Taeuber and Rosenwaike, 1992), an important implication of our findings
relates to the long-term care for the disabled elderly, who in 1990 represented more
than half of the nation's elderly population (Torres-Gil, 1996). As friends and neigh­
bors are not suitable substitutes for adult children as the major providers of long-term
care (Litwak, 1985), our findings suggest it is viable that adult children will remain the
main providers of long-term care for the elderly. However, policymakers should realize
that long-term care tends to be extremely burdensome (Brody, 1985). If formal institu­
tions are not set up to relieve part of this burden, the attraction of the elderly to their
adult children may be weakened (Anderson, 1977), leading to the degeneration of
the MEF system and the need for further expansion of long-term care facilities for the
elderly.
Acknowledgements. The research is supported by the grants from the NIA (grant number ROI­
AGI2291), the NICHD (grant number ROl-HD297525), and the Social Sciences and Humanities
Research Council of Canada (grant number 412-98-0008). The migration data for this paper were
prepared at the Population Studies Center, University of Michigan, from the 1990 US Census files.
The authors acknowledge Cathy Sun for computer programming assistance. We are grateful to the
helpful comments of the referees and to Eugene Litwak's help in improving our knowledge on US
families. The remaining flaws are ours.
References
Anderson M, 1977, "The impact on family relationships of the elderly of changes since Victorian
times in governmental income-maintenance provision", in Family, Bureaucracy, and the
Elderly Eds E Shanas, M B Sussman (Duke University Press, Durham, Nq pp 36 - 59
Biblarz T J, Raftery A E, 1999, "Family structure, educational attainment, and scoioeconomic
success: rethinking the 'pathology of matriarchy'" American Journal of Sociology 105 321 - 365
Bradsher J E, Longino C F, Jackson D J, Zimmerman R S, 1992, "Health and geographical
mobility among the recently widowed" Journal of Gerontology 47 S261- S268
Brody E M, 1985, "Parent care as a normative family stress" The Gerontologist 2519-29
Bultena G L, Marshall D G, 1970, "Family patterns of migrant and nonmigrant retirees" Journal
of Marriage and the Family 32 89 - 93
Clark D E, Knapp T A, White N E, 1996, "Personal and location-specific characteristics and
elderly interstate migration" Growth and Change 27 327 - 351
(19) The Million Man March, spearheaded by Minister Louis Farrakhan, a black Muslim, took
place on 16 October 1995 in Washington, DC. It was attended by numerous black men who were
urged to pledge to be, among other things, good fathers and good husbands (Madhubuti and
Karenga, 1996). It reflected a clear recognition by major black leaders of the need for strengthening
the black family system in the United States.
(20) It is encouraging that we learned recently (September 2000) via an e-mail ([email protected])
from the NSFH Office at the University of Wisconsin (www.ssc.wisc.edu/nsfh/) that geographical
codes will be merged to the individual records of the NSFH. We plan to continue this line of
research as soon as the merged data become available.
, .
214
~f­
:~
K·L Liaw, W H Frey, J.p Lin
Clark R L, Wolf D A, 1992, "Proximity of children and elderly migration", in Elderly Migration
and Population Distribution Ed. A Rogers (Belhaven Press, London) pp 77 - 96
Cribier F, 1980, "A European assessment of aged migration" Research on Aging 2 225-270
Eldridge H T, 1965, "Primary, secondary, and return migration in the United States, 1955 - 60"
Demography 2 444-455
Franklin D, 1997 Ensuring Inequality: The Structural Transformation ofthe African·American
Family (Oxford University Press, New York)
Frazier E F, 1939 The Negro Family in the United States (University of Chicago Press, Chicago, IL)
Frey W H, Liaw K·L, Xie Y, Carlson M J, 1996, "Interstate migration of the US poverty population:
immigration 'pushes' and welfare 'puUs' Population and Environment 17491 - 536
Frey W H, Liaw K·L, Lin G, 2000, "State magnets for different elderly 'migrant types' in the
United States" International Journal ofPopulation Geography 6 21-44
Glasgow N, Reeder J, 1990, "Economic and fiscal implications of non metropolitan retirement
migration" Journal ofApplied Gerontology 9 433 - 451
Haas III W H, Serow W J, 1993, ''Amenity retirement migration process: a model and preliminary
evidence" The Gerontologist 33 212 - 220
HCFA, 1990 Medicare and Medicaid Data Book publication 03314, Health Care Financing
Administration, US Department of Health and Human Services, 7500 Security Boulevard,
Baltimore, MD 21244·1850
Hogan D P, Eggebeen D J, Clogg C C, 1993, "The structure of intergenerational exchanges in
American families" American Journal ofSociology 98 1428 - 1458
Jaynes G D, Williams Jr R M (Eds), 1989 A Common Destiny: Blacks and American Society (National
Academy Press, Washington, DC)
Kanaroglou P, Liaw K-L, Papageorgiou Y Y, 1986, ''An analysis of migratory systems: 2.
Operational framework" Environment and Planning A 181039-1060
Liaw K-L, Ledent J, 1987, "Nested logit model and maximum quasi-likelihood method: a flexible
methodology for analyzing interregional migration patterns" Regional Science and Urban
Economics 17 67 88
Lichter D T, Eggebeen D J, 1993, "Rich kids, poor kids: changing income inequality among
American children" Social Forces 71 761 780
Lin G, Rogerson P A, 1995, "Elderly parents and the geographical availability of their adult
children" Research on Aging 17 303 - 331
Litwak E, 1960a, "Occupational mobility and extended family cohesion" American Sociological
Review 25 9 - 21
Litwak E, 1960b, "Geographical mobility and extended family cohesion" American Sociological
Review 25 385 394
Litwak E, 1965, "Extended kin relations in an industrial democratic society", in Social Structure
and the Family: Generational Relations Eds E Shanas, G F Streib (Prentice-Hall, Englewood
Cliffs, NJ) pp 290- 323
Litwak E, 1985 Helping the Elderly (Guilford Press, New York)
Litwak E, Longino C F, 1987, "Migration patterns among the elderly: a developmental perspective"
The Gerontologist 27 266 - 272
Long L, 1988 Migration and Residential Mobility in the United States (Russell Sage Foundation,
New York)
Longino C F, 1979, "Going home: aged return migration in the United States, 1965-1970"
Journal of Gerontology 34 736 - 745
Longino C F, 1995 Retirement Migration in America (Vacation Publications, Houston, TX)
Longino C F, Serow W J, 1992, "Regional differences in the characteristics of elderly return
migration" Journal of Gerontology 47 S38 - S43
McCullagh P, 1983, "Quasi-likelihood functions" The Annals ofStatistics 11 59 - 67
McFadden D, 1974, "Conditionallogit analysis of qualitative choice behavior", in Frontiers in
Econometrics Ed. P Zarembka (Academic Press, New York) pp 105 142
MacMahon W W, Chang S C, 1991, "Geographical cost ofliving differences: interstate and
intrastate update 1991 ", MacArthur/Spencer Series, number 20, Center for the Study of
Educational Finance, Illinois State University, Normal, IL
Madhubuti H R, Karenga M (Eds), 1996 Million Man March/Day ofAbsence (Third World Press,
Chicago, IL)
Miller A R, 1977, "Interstate migrants in the United States: some social-economic differences by
type of move" Demography 141-17
\
.
'.
Adult children and elderly migration
215
Morgan S P, McDaniel A, Miller A T, Preston S H, 1993, "Racial differences in household
structure at the turn of the century" American Sociological Review 98 798 - 828
Morrison P A, 1990, "Demographic factors reshaping ties to family and place" Research on Aging
12399-408
Moynihan D P, 1965, "The negro family: the case for national action", US Department of Labor,
Washington, DC
Newbold K B, 1996, "Determinants of elderly interstate migration in the United States, 1985 - 90"
Research on Aging 18 451 - 476
Parsons T, 1949, "The social structure of the family", in The Family: Its Function and Destiny
Ed. R N Anshen (Harper & Row, New York) pp 241 - i75 [revised edition 1959]
Pezzen L E, Schone B S, 1999, "Parental marital disruption and intergenerational transfers:
an analysis of lone elderly parents and their children" Demography 36 287 - 297
Rogers A, 1990, "Return migration to region of birth among retirement-age persons in the
United States" Journal of Gerontology: Social Sciences 45 SI28 - S134
Rogerson P A, Burr J A, Lin G, 1997, "Changes in geographical proximity between parents and
their adult children" International Journal of Population Geography 3 121 - 136
Rowles G D, 1983, "Between worlds: a relocation dilemma for the Appalachian elderly"
International Journal of Aging and Human Development 17 301 - 314
Ruggles S, 1994, "The origins of African-American family structure" American Sociological
Review 59 136 - 151
Serow W J, 1978, "Return migration of the elderly in the USA: 1955 - 1960 and 1965 - 1970"
Journal of Gerontology 33 288 - 295
Silverstein M, 1995, "Stability and change in temporal distance between the elderly and their
adult children" Demography 32 29 - 45
Silverstein M, Bengston V L, 1997, "Intergenerational solidarity and the structure of adult child­
parent relationships in American families" American Journal ofSociology 103 429 - 460
Smith H L, Morgan S P, Koropeckyj-Cox T, 1996, "A decomposition of trends in the nonmarital
fertility ratios of blacks and whites in the United States, 1960 - 1992" Demography 33 141 - 151
Speare A Jr, McNally J, 1992, "The relation of migration and household change among elderly
persons", in Elderly Migration and Population Distribution Ed. A Rogers (Belhaven Press,
London) pp 61- 76
Sussman M B, 1965, "Relationships of adult children with their parents in the United States", in
Social Structure and the Family: Generational Relations Eds E Shanas, G F Streib (Prentice­
Hall, Englewood Cliffs, NJ) pp 62-92
Taeuber C M, Rosenwaike I, 1992, "A demographic portrait of America's oldest old", in The
Oldest Old Eds R M Suzman, D P Willis, K G Manton (Oxford University Press, Oxford)
pp 16-49
Torres-Gil F M, 1996, "Forward: health care reform, long-term care, and the future of an aging
society", in From Nursing Homes to Home Care Eds M E Cowart, J Quadagno (Haworth
Press, New York) pp xiii-xviii
Townsend P, 1957 The Family Life of Old People: An Enquiry in East London (Routledge and
Kegan Paul, London)
USDC, 1991 State and Metropolitan Area Data Book 1991 US Department of Commerce,
Superintendent of Documents (US Government Printing Office, Washington, DC)
Wilson W J, 1987 The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy
(University of Chicago Press, Chicago, IL)
Wirth L, 1938, "Urbanism as a way of life" American Journal ofSociology XLIV I - 24
Young M, Willmott P, 1957 Family and Kinship in East London (The Free Press of Glencoe, IL)
I.
•
K-L Liaw, W H Frey, J-P Lin
216
APPENDIX Definitions of the place attributes used as covariates in the nested logit model In assessing the effects of adult children and environmental amenities on the interstate
migration behavior of native elderly black people and white people, we control for
the effects of other place attributes by including them as covariates in the nested
logh model. These covariates are defined as follows. The data sources that are not
specifically identified below are indicated in Freyet al (1996).
Cost-ol-living index: the cost of living in a given state in 1985, with the national average
set at 100. Data source: MacMahon and Chang (1991).
Medicare: the 1987 Medicare payment per elderly recipient. The unit is US $1000 per
person. Data source: USDC (1991, page 220).
Medicaid: the 1986 Medicaid payment per elderly recipient. The unit is US $1000 per
person. The missing value of Arizona is replaced by the average of the other states.
Data source: HCFA (1990).
Income: the income per capita of a state computed in the following way: first, adjust
the state-specific 1985 and 1989 nominal per capita incomes by the corresponding
state-specific cost of living indices of the same years; second, the 1985-adjusted and
1989-adjusted values are then averaged. The unit is US $10 000 per person.
Employment growth: for each state, this variable is the state-specific 1985-89 growth
of total civilian employment divided by the 1985 total civilian employment. The unit is
'proportion per four years'.
In(distance}: the natural logarithm of the popUlation gravity centers of origin and
destination states. It is calculated as In(distancelmiles).
Contiguity: for each potential destination, this is a dummy variable assuming the value
of 1 if the destination shares a common border with the state of origin.
'Racial'similarity: for the migrants of a specific 'race' in the destination choice sub­
model, this is the logit of the specific 'race's' proportional share of the population of
the potential destination in 1985, computed indirectly from the data of the 1990 Census.
For the potential migrants of a specific 'race' in the departure submodel, this is the
logit of the specific 'race's' proportional share of the population in the origin in 1985,
computed indirectly from the data of the 1990 Census.
Proportion homeowners: this variable is the proportion (percentage) of the elderly
(aged 65 or more years) owning homes in 1990. Data source 1990 Census 5% PUMS.
In (population size}: the natural logarithm of the population size of the state in 1985,
computed indirectly from the data of the 1990 Census. It is calculated as In(popula­
tionllOOO 000 persons). Industrial heartland: a dummy variable assuming the value of 1 if the 1985 state of residence is Delaware, Maryland, or Washington, DC, or is in the Middle Atlantic Division or the East North Central Division. p
© 2002 a Pion pUblication printed in Great Britain
\
.